New or changed in 0.4.0:
- The
showCompiledoption lets you show compiled code in error messages- Code output in error messages now includes source line numbers
- You can specify what
moduleto use for the various language engines- The
printResultsoption now defaults tofalse, as it tends to produce unwanted output in typical usage.
What better place to specify your code's behavior than in its API documentation? And how better to document your API than with examples? But if they're not tested, examples tend to get stale and out-of-date. And testing them by hand is a pain.
But what if you could automatically test your examples, as part of your project's existing mocha test suites? For that matter, what if you could do documentation-first testing, by writing the API docs for the API you're creating, and using the examples in them as tests to drive your initial development?
Mockdown lets you do all of these things, and more, by testing code samples embedded in markdown files, like this:
// "Hello world" Sample
console.log("Hello world!")Hello world!
The above is a documentation test, or "doctest". You embed a code block to be run, optionally followed by a blockquoted code block representing its output. (If you don't include the output block, it's the same as asserting the example will output nothing.)
If the output doesn't match, or an unexpected error is thrown, the test fails. If your test completes asynchronously, you can use wait() to defer the test's completion until a callback, promise resolution, or other asynchronous result occurs:
// Using wait() with setTimeout()
var done = wait(); // wait() with no arguments returns a callback... but we
// could have given it a promise, predicate, or timeout
// to wait for instead
setTimeout(function(){
console.log("Hello world!");
done();
}, 50);Hello world!
Section headings in your markdown files define mocha suites, so your test suites will precisely match the table of contents of your documentation files. If you need to do things like mark tests to be skipped or ignored, you can add simple HTML comment directives like this:
<!-- mockdown: ++skip -->in order to mark a test pending, override the language defaults, change how output is checked, etc.
And since these directives are HTML comments, they don't show up when viewing the docs on github, or in any HTML docs you're generating for your website. (Which means your readers won't get distracted by stuff that only matters to your testing process.)
Mockdown was inspired and influenced both by Python's doctest module and Ian Bicking's DoctestJS, but is a new implementation specifically created to work with mocha and markdown. Unlike DoctestJS, it:
- Works with markdown on the server instead of HTML in the browser
- Uses standard Node console inspection utilities instead of rolling its own pretty-print facilities, and
- Supports other languages besides plain Javascript -- including the use of multiple languages in the same document (Babel and CoffeeScript out of the box, but you can supply your own engine(s) via the options)
- Uses mocha for test running and reporting, allowing integration into existing test suites
For the most part, tests are free-form. Just insert code blocks, optionally followed by blockquoted code blocks (> -prefaced) to specify expected output. Markdown section headings delineate mocha suites, with the heading levels being used to determine the nesting structure, and the section titles used to name the suites.
By default, code is assumed to be plain Javascript that can be executed by the engine mockdown is running on, unless you use a fenced code block with an explicit language declaration. (You can change this default, add language engines, etc. using directives and options, which are discussed in later sections below.)
Both the sample code and the expected output can be either Github-style "fenced" code blocks or traditional 4-space indented code blocks. Remember, however, that blockquotes require a space after the >, so if you are using indented code blocks for expected output, you will need 5 spaces between the > and the beginning of expected output. That is:
> this is not an expected output block!> but this is! If you don't include all five spaces, the markdown parser will see the blockquote as regular text, so mockdown won't see it as an expected-output block, and your test will fail, despite it looking okay to the naked eye.
If your code sample's first non-blank line begins with a line comment (//, #, --, or %, depending on your language), it'll be used as the test's title, e.g.:
// This will show up as the mocha test title
{ /* We're not testing anything here except the test title */ }Or, if the code sample doesn't start with a line comment, but it's the only code sample within a given suite, it'll replace the suite and assume its title. (In other words, if you have only one code sample and no subheadings under a given markdown heading, then the test will get its title from the heading.)
If there's more than one code sample under a given heading, or if there are subheadings under that heading, then any otherwise-untitled tests will be titled "Example N at line M", where N is its sequence number within its suite, and M is the line number of its first code line.
Code samples are run in a virtual environment with a simulated console, using the mock-globals module. Anything that a code sample outputs via console.log, console.error, etc. will be sent to the console's output record, which is then compared against the expected output at the end of each code sample's execution.
If the output doesn't match the expected output, the corresponding test will fail with a detailed error message showing the actual and expected output (unless you suppress it by changing the relevant options.)
By default, the last value evaluated in a code sample is printed, in much the same way as the Node REPL, with undefined results remaining silent. You can change this behavior, however, using the printResults and ignoreUndefined options. (Either by passing different options to the API, or by using directives, as will be described in later sections below.)
(Note: mock-globals is not a secure execution environment. Do not use mockdown to process files from untrusted sources, or you will be very sorry!)
You can include error output in your samples, if the purpose of the example is to show an error. Only the error message itself will be printed to the virtual console, unless you have a non-zero stackDepth option set (via the API or an in-document directive).
So, the following example throws an error, but since an error is the intended result, the test will be still considered successful:
throw new Error("this is the message")Error: this is the message
If the error name or message had differed, the test would fail instead, and the error stack shown by mocha would be the stack from the first error issued by the code sample.
If your code sample completes asynchronously, you need to use the wait() function to defer the test's output matching until your code is finished running.
When called without any arguments, wait() returns a node-style callback that can be invoked to finish the test. If you call it with an error, the error will be written to the virtual console for output matching purposes. So if the error is expected, the test will still succeed.
If you are working with promises instead, you can call wait(aPromise) to make the test wait for the promise to finish. As with the callback scenario, a promise rejection is treated as an error that gets written to the virtual console for output matching purposes.
If you have neither a callback-taking function or a promise, you can still use wait(timeout) to wait the specified number of milliseconds, wait(aFunction) to call aFunction() every millisecond until it returns true, or wait(interval, aFunction) to do the same thing with a specified number of milliseconds between checks.
(By the way, if you need wait() to have a different name because you need to use the name wait for something else in your examples, you can rename it by changing the waitName option in a directive, or in the options you supply to the mockdown API. See the sections below for more details.)
For some documentation, including an explicit wait() call may be intrusive; for these situations, you can use the waitForOutput option in a directive. For example, the following test is configured to wait for the string "done" to appear in the output, using the directive <!-- mockdown: waitForOutput = "done" -->:
setTimeout(function(){ console.log("done"); }, 10);done
(The main downside to this approach is that if the desired string never appears in the output, the test will time out before any output comparison is done.)
For more on how to use directives to configure options, see the next few sections.
Sometimes, you need to have mocha skip a test and mark it pending. Other times, you may have code blocks in your documentation that you don't want to treat as tests at all! You can use directives to control these things, as well as to set other options like waitName or printResults. A directive is a special kind of HTML comment, set off by blank lines before and after it. For example, to skip a single test and mark it pending, you can use:
<!-- mockdown: ++skip -->Or to treat the next code block as a non-test code block, you can use:
<!-- mockdown: ++ignore -->If you want to mark multiple tests to skip or code blocks to ignore, you can bracket them with a pair of mockdown-set directives, like so:
<!-- mockdown-set: ++skip -->
Tests between these directives will be marked
"pending" in mocha!
<!-- mockdown-set: --skip -->The main difference between a mockdown directive and a mockdown-set one is that mockdown only affects the next code block encountered, while mockdown-set changes the current default options for the document.
So, anything you set with mockdown-set will stay set, until you change it with another mockdown-set -- even if it's temporarily overridden for one test with a mockdown directive. (This makes it easy to change an option for just one test, because you don't have to remember to change things back afterwards: just use a mockdown directive for anything that should apply to just one test.)
(One other important difference between mockdown and mockdown-set directives is that mockdown directives must appear immediately before the code blocks they affect, without any other text or non-mockdown directives in between. Otherwise, parsing of the document will fail with a SyntaxError, to avoid any ambiguity as to the intended effects.)
Directives aren't limited to toggling boolean flags like skip and ignore. You can also set non-boolean options values (e.g. <!-- mockdown: stackDepth = 3 -->) and even combine multiple option changes in a single directive, e.g.:
<!-- mockdown: stackDepth=3; waitName="defer"; ++showDiff -->If this looks a lot like Javascript code, that's because it is! Directive bodies are actually code that runs in a separate mock-globals environment, where global variables are tied directly to the options for the next code block (in a mockdown directive), or the document defaults (in a mockdown-set directive).
When used with a boolean option, the ++ and -- operators are shorthand for setting the option to true or false, respectively. This happens even if you use them repeatedly, so you don't need to keep track of how many times you incremented or decremented them: ++ always turns the option on, and -- always turns it off.
Although directives are Javascript code, it's important to understand that this code runs while your markdown document is being parsed, not while the tests are running. So they can only configure tests, not intervene in their execution.
For this reason, there is also a third type of directive: mockdown-setup. You can use this directive at most once in a given markdown file, and only before any other directives or code blocks appear in the file. Within this directive, your Javascript code can set or change the globals and languages that will be used by your tests. It is in all other respects identical to a mockdown-set directive (i.e., you can use it to set other defaults for the file).
There are a great many options you can set or change via directives or the mockdown API; the next section lists them all.
All of the options described in this section can be set or changed within a markdown document using directives (as described in the previous section). They can also be passed in as options to mockdown.testFiles() and other mockdown APIs (as described in the section on "Running Your Tests", below).
Object, default: {}.
An object containing the pseudo-global variables that will initialize the mock-globals Environment where the code samples will execute. Can only be configured via the options passed into mockdown's APIs, or via a mockdown-setup directive at the top of a file. (That is, unlike other options, it can't be set via mockdown or mockdown-set directives.)
String or null/undefined, default: "wait"
The name the wait() function is made available under. You can change this in order to avoid conflict with a name in your examples. If this option is set to null or undefined, the wait() function will not be accessible from the example.
String or null/undefined, default: "test"
The name the current mocha test object will be made available under, so you can e.g. change the test timeout. If this option is set to null or undefined, the test object will not be accessible from the example.
Boolean, default: false
If true, mark the applicable test(s) pending in Mocha.
Boolean, default: false
If true, do not turn markdown code blocks into examples until it becomes false again. Can be used with a mockdown directive to "comment out" a single test, or a pair of mockdown-set directives to comment out a group of tests.
Boolean, default: true
When a test fails due to unmatched output, show the full expected and received output, along with the source code of the test.
Boolean, default: false (New in version 0.4.0)
When a test fails, show the compiled code in the error output, instead of the original source. (Has no effect if showOutput is false.) This can be useful for debugging a broken test case or example written in a compile-to-JS language.
Boolean, default: false
When a test fails due to unmatched output, tell mocha to diff the output.
Integer from 0 to Infinity (i.e., unlimited stack depth). Default: 0.
When a test throws an unhandled exception, how many lines of stack trace should be included in the output? (This lets you add more lines temporarily for debugging, or permanently if the contents of the stack trace are what your example is testing.)
Optional function, string, or regular expression predicate; default undefined
If set, the test is considered asynchronous (as if wait() were explicitly called), and each string written to the virtual console will be checked using the supplied function, string, or regular expression. If there is a match, the test will end on the next process tick, without needing to explicitly call a done() function.
A function predicate is considered to match if it returns a truthy value for a given output string, and a string predicate is considered to match if it is found at any point within the output string. Regular expression predicates are matched with .match() on the output string.
If a matching string is never written, the test will time out, and then the output up to that point will be compared with the expected output.
Boolean, default: false. (CHANGED in 0.4.0)
If true, the virtual environment acts like the node REPL, printing the value of the last expression in a code sample.
For example, the following test requires printResults to be true:
6 * 742
Boolean, default: true.
If true, don't print an undefined result. This only has any effect if printResults is true.
Function or undefined, default: undefined
The function used to convert a REPL result to a string suitable for writing. Only has effect if printResults is true. If writer is undefined, then the Node repl module's current writer property will be used (which by default is a slightly modified version of util.inspect()).
String, default "javascript"
The name of the language that should be used when a code block doesn't have an explicit language. That is, the language to be used for indented code blocks, and fenced blocks without a specified language. You can also set this to "ignore" to ignore such code blocks and not create tests for them.
Language names are case-insensitive, at least in the sense that they are converted toLowerCase() before being looked up in the languages mapping.
Object, default: require('mockdown/languages')()
An object whose keys are all-lowercase language names, and whose values are language aliases or language engines. A language alias is just a string that names the engine to be used, so for example if you set languages.es7 = "babel", this would tell mockdown to use Babel to compile code blocks with a language of es7.
(Note: aliases are not recursive; they must name a language engine, not another alias. They can, however, be set to "ignore", which indicates blocks of that language should be ignored and not used as tests.)
A language engine is an object with one required property, toJS:, which must be a function accepting a mockdown.Example object and a starting line number, and returning a string of Javascript. Usually, language engines will also include an options: property that will be used to send compiler options to the underlying language. (For example, you can set languages.babel.options.stage to change the stability level used for Babel examples.)
Note that the languages option can only be configured via the options passed into mockdown's APIs, or via a mockdown-setup directive at the top of a file. (That is, like globals, it can't be set via mockdown or mockdown-set directives.)
Currently, the default mapping for this option includes language engines for:
babel(withes6as an alias)coffee(withcoffee-scriptandcoffeescriptas aliases)javascript(withjsas an alias)
And it includes html, markdown, and text as aliases for ignore.
You can add your own aliases and engines to this mapping by in-place modification in mockdown-setup code, or by passing a replacement object as a languages: option to the API. (You can also call require('mockdown/languages')() to get a copy of the defaults that you can then modify and pass in.)
Both the babel and coffee-script language engines have a module property that can be used to determine what module to load to do the compiling. languages.babel.module defaults to "@babel/core", and languages.coffee.module defaults to "coffeescript", but you can override them if you need to.
The filename that will appear in stack traces for errors thrown by code or directives within the file. Like globals and languages, it can't be changed on the fly, but only initialized by passing options to the API or in a mockdown-setup directive. If you don't explicitly provide it to the API, and mockdown loads the file for you, it will be set to the filename it was asked to load. String, defaults to "<anonymous>" if the document was parsed from a string instead of a file.
To include documentation files in your test suites, just pass a list of filenames, the mocha suite/describe and test/it functions, and an optional options object to mockdown.testFiles(), like this:
var mockdown = require('mockdown');
mockdown.testFiles(['README.md'], describe, it, {
printResults: false, // disable REPL mode
globals: {
// supply some global vars for your code samples
someUsefulFunction: function () {
// your examples will now be able to call
// `someUsefulFunction()` without needing
// to `require()` it
return "I'm useful!";
},
// you can mock or stub any global names, too!
require: function(path) {
if (path === 'mymodule') return require('./');
else return require(path);
}
}
})As you can see above, the options argument not only lets you set any mockdown options or globals, it also lets you access non-virtualized code. If you configure your globals from inside a mockdown-setup directive, you only have access to the virtual environment where directives run. But when you configure them via the API, you can use functions that have access to e.g. the "real" require() function.
If you call mockdown.testFiles() from the top-level code of a module, the added test suites will be at the top level of your overall test set. If you call it from inside a suite or describe() block, the suites will be nested within that block.
Alternately, if you want more explicit control over the process, you can:
-
Create a parser object using
parser = new mockdown.Parser(options) -
Get a
mockdown.Documentusingdoc = parser.parse(text)ordoc = parser.parseFile(path), and -
Register tests and suites with mocha by calling
doc.register(suiteFn, testFn)withdescribeanditor their equivalents in the mocha interface you're using.The
.register()method can optionally be given a mock-globalsEnvironmentobject as a third parameter, in which case it will be used instead of creating a new one. (But in that event, theoptions.globalswon't be used; you'll have to configure theEnvironmentinstance yourself.)
If you're parsing strings, you'll probably want to include a filename: entry in the options you give the parser, so that error messages will include the right filename. And for a complete list of all the options you can use with any of mockdown's APIs, see the section on "Configuring Your Tests", above.
Currently, multi-language support is still experimental. Most JS transpilers expect to be producing an entire module at a time, rather than a collection of code fragments to be run REPL-style. So although it "works", you may run into language-specific compilation issues, and may need to use the showCompiled option to display compiled source in error messages so you can see what's really going on.
(Also, note that although mockdown includes engines for Babel and CoffeeScript, it does not declare dependencies on them, as by design it should use your installation of the relevant compiler modules. That way, it will always be in sync with your project's version of the associated compiler.)
- The
waitForOutputoption lets you match console output to end an asynchronous test, without needing to callwait()in the test itself. See the section above on asynchronous tests for an example.
- Multi-language support is still experimental, and there aren't any docs yet on how to create an engine correctly.
- No API docs except nearly 2000 lines of very verbose tests
- Since it hasn't really been used yet, there are probably lots of syntax corner cases that haven't been encountered yet
- Ellipsis and whitespace options for output matching aren't implemented